Empirical evidence indicates that students who adopt healthier lifestyles tend to experience higher levels of subjective well-being, even after accounting for sociodemographic differences (Amiri et al., 2019, 2023). These findings suggest that health-promoting behaviors—such as balanced nutrition, regular physical activity, and effective stress management—exert a sustained protective influence on mental health and overall life satisfaction. Thus, a health-promoting lifestyle emerges as a critical correlate of well-being among university students, reinforcing its role as a foundational component of adaptive functioning and positive development in academic contexts (Lee and Loke, 2005).

On the other hand, research has shown that a health-promoting lifestyle is associated with lower levels of depressive and anxiety symptoms among university populations. A recent study conducted with Honduran students reported that greater adherence to healthy behaviors was significantly associated with lower levels of psychological distress, acting as a protective factor against emotional discomfort (Castillo-Díaz et al., 2024). This finding aligns with evidence from Chinese university students, where health-promoting lifestyles have been found to correlate negatively with depression, even after adjusting for social support and sociodemographic characteristics (Tang et al., 2021). Furthermore, previous research has documented that physical activity—a core component of a health-promoting lifestyle—is associated with significant reductions in anxiety and depressive symptoms, and with higher levels of psychological well-being (Buecker et al., 2021; Herbert, 2022), highlighting both preventive and therapeutic benefits.

Similarly, the regular practice of health-promoting behaviors—such as engaging in physical activity, maintaining a balanced diet, and adopting self-care routines—has been associated with greater stability of circadian rhythms and a more favorable subjective perception of nighttime rest (Zhang et al., 2023). Consistent with this, recent research has found that students who report higher levels of health-promoting behaviors tend to exhibit more favorable sleep patterns and a lower incidence of fatigue or daytime sleepiness (Alothman et al., 2024; Sun et al., 2024).

Higher levels of anxiety have been found to be associated with lower scores of subjective well-being, indicating an inverse relationship between the two variables (Liu et al., 2009; Zhu et al., 2023; Benítez-Agudelo et al., 2025). Similarly, studies conducted among university populations have reported that the presence of depressive symptoms is linked to significantly lower levels of subjective well-being (Liu et al., 2009; Rossi et al., 2019). Additional findings from European student samples confirm that mental well-being decreases in the presence of elevated indicators of emotional distress, further reinforcing this pattern (Navarra-Ventura et al., 2024).

On the other hand, sleep quality has been positively associated with subjective well-being, with evidence showing that better perceived sleep quality corresponds to higher levels of well-being (Su and He, 2023). Intraindividual temporal analyses among university students have revealed concurrent associations between daily variations in sleep and well-being (McGowan et al., 2023). Moreover, findings indicate that from the very first semester of university, shorter sleep duration and greater variability in sleep patterns are linked to poorer indicators of well-being (Bono and Hill, 2022). Taken together, these findings support the role of mental health and sleep quality as key pathways through which psychological functioning and daily experiences shape students’ well-being. However, the complexity of these interrelations remains only partially understood, pointing to the need for more integrated and context-sensitive analyses.

Despite the arguments presented above, the interplay among health-promoting behaviors, mental health, sleep quality, and well-being in university populations remains insufficiently understood. Moreover, most research on well-being has been conducted in high-income countries, typically involving populations whose structural, cultural, and economic conditions differ substantially from those of Latin American contexts (Thomas et al., 2021; Sollis et al., 2024; Roy et al., 2025). Considering these factors, it is essential to generate context-specific evidence on the factors that may contribute to student well-being.

In response to this gap, the present study aims to evaluate a structural equation model (SEM) in which a health-promoting lifestyle is specified as a predictor of well-being, with affective symptomatology (depression and anxiety) and sleep quality included as mediating variables among Honduran university students. Building on the empirical evidence and conceptual framework outlined, this model examines both direct and indirect pathways to well-being, addressing the limitations of prior research that has often considered these variables in isolation. The SEM approach enables the combine estimation of direct and indirect effects among observed and latent variables (Beran and Violato, 2010), offering a more comprehensive understanding of the behavioral, psychological, and physiological mechanisms underlying well-being.

Figure 1 depicts the hypothesized structural model. Based on prior evidence, health-promoting lifestyle is expected to positively predict well-being and sleep quality and negatively predict anxiety and depressive symptoms. In turn, anxiety and depressive symptoms are expected to negatively predict well-being, whereas sleep quality is expected to positively predict it. With regard to the mediational pathways, the following hypotheses are proposed:

The proposed approach is particularly relevant in contemporary psychological and health research, enabling the examination of complex relationships among variables by integrating behavioral, cognitive, and emotional dimensions. The findings may serve as valuable input for identifying patterns of association that can inform psychoeducational interventions and institutional policies aimed at promoting comprehensive health and enhancing the well-being of higher education students.

This study employed a non-experimental, analytical cross-sectional design. It was conducted in a higher education context at a Honduran public macro-university with nationwide coverage.

The population of the study was composed of Honduran university students from the 2024 and 2025 first-year cohorts, which together comprised 22,850 students (13,379 in 2024 and 9,471 in the first academic term of 2025). Through a non-probabilistic sampling strategy, a total of 6,825 students participated voluntarily; however, 121 were excluded for not meeting the inclusion criteria, which required being a first-year student, aged 18 years or older, enrolled in any academic discipline offered by the university across all campuses nationwide, and providing attentive and non-random responses. Further details on data quality control are described in the Procedures section. The final sample in this study consisted of 6,704 university students, representing 29.3% of the total first-year student population. Sociodemographic characteristics of the sample are summarized in Table 1.

Data collection was conducted between January 2024 and June 2025 through online self-administered questionnaires distributed via institutional communication channels, academic events, and classroom spaces regularly attended by first-year university students. The dataset used in this study forms part of a larger research project entitled “Psychosocial determinants of well-being and quality of life in university students: a diagnostic and longitudinal evaluation.” All study procedures strictly adhered to the principles of the Declaration of Helsinki and local legislation. The larger research project received ethical approval from the Ethics Committee of the Faculty of Social Sciences at the National Autonomous University of Honduras (Ref. CEIFCS-2024-P01). Participation was entirely voluntary, and digital informed consent was required prior to enrollment.

To ensure the methodological integrity and quality of the data, several quality control procedures were implemented. The online form included embedded attention check items (e.g., “Select option 4 for this question”) and response pattern screening to identify and exclude cases exhibiting straightlining or repetitive response behaviors across multiple scales (e.g., consistently selecting the same option throughout the questionnaire). In addition, minimum completion time thresholds were applied to detect inattentive or automated responses. Specifically, responses completed in less than 15 min—equivalent to approximately 30% of the mean completion time of 49 min observed during the pilot and initial data collection—were flagged for removal. The combined use of attention checks, response pattern detection, and response time screening follows best practices for identifying insufficient effort responding and improving data validity in online questionnaire research (Brühlmann et al., 2020).

As a preliminary step, and considering that all study variables were collected through self-report measures at a single time point, the potential presence of common method bias was examined. To this end, Harman’s single-factor test was conducted by means of an exploratory factor analysis (EFA) including all items from the instruments employed in this study. Following conventional criteria, if the first factor accounted for less than 40% of the total variance, the presence of a severe common method bias was deemed unlikely (Podsakoff et al., 2003).

Descriptive statistics for all variables were computed, including means, standard deviations, minimum and maximum values, histograms, skewness, and kurtosis. Subsequently, bivariate correlations among the main study variables were estimated using the factor scores derived from the measurement models of each instrument (HPLP-II, PHQ-9, GAD-7, and WHO-5), along with the observed score from the SQS. In the case of the HPLP-II, only the correlation for the general factor was reported, as the hierarchical model conceptualizes first-order factors as indicators of the overarching construct.

To test the study hypotheses, a structural equation modeling (SEM) approach was employed. SEM simultaneously integrates measurement models—where latent variables account for their respective observed indicators—and a structural model that estimates the direct and indirect relationships among latent constructs. The measurement models for each latent variable (health-promoting lifestyle, anxiety, depression, and well-being) were assessed through item-level confirmatory factor analyses (CFA), with detailed results reported in the instrument descriptions. Given that sleep quality was represented by a single item, it was analyzed using its observed scores. The conventional sequence of steps for conducting structural equation modeling—specification, identification, estimation, evaluation, and modification—was implemented in accordance with established methodological literature (Kline, 2023).

Figure 1 depicts the tested structural model. The specification of the model indicates a multiple mediation structure in which health-promoting lifestyle predicts well-being both directly and indirectly through depressive symptoms, anxiety symptoms, and sleep quality. The weighted least squares mean- and variance-adjusted (WLSMV) estimator was used for model estimation (Schumacker and Lomax, 2015). Model fit was evaluated using the following indices: comparative fit index (CFI) and Tucker–Lewis index (TLI) values ≥ 0.90 were considered acceptable and ≥ 0.95 indicative of good fit; standardized root mean square residual (SRMR) and root mean square error of approximation (RMSEA) values ≤ 0.08 were regarded as acceptable and ≤ 0.06 as evidence of good fit (Hu and Bentler, 1999; Marsh et al., 2004). We examined direct effects on each endogenous variable, specific indirect effects (via each mediator), total indirect effects (summed across mediators), and overall effects (direct plus indirect). In addition, the proportion of the total effect mediated by each path was computed to facilitate interpretation. Given the large sample size, the delta method was deemed appropriate for computing 95% confidence intervals for all structural effects (Sim et al., 2022).

All analyses were conducted in R (version 4.4.2) using the RStudio environment (version 2025.5.1.513). The skimr package (version 2.1.5) (Waring et al., 2022) was used for descriptive analyses, whereas semTools (version 0.5–6) (Jorgensen et al., 2022) and lavaan (version 0.6–16) (Rosseel, 2012) were employed for the estimation of measurement and structural models.

As part of the preliminary analyses, common method bias was examined. The EFA results indicated that eight factors had eigenvalues greater than 1, and the variance explained by the first factor was 28.49%, which is below the critical threshold of 40%. These findings suggest that no severe common method bias was present in the data.

Table 2 presents the descriptive statistics for the variables included in the study. Skewness and kurtosis values fell within the acceptable range (−2 to 2), indicating no severe deviations from normality (Mishra et al., 2019). Regarding health-promoting lifestyle, scores on the HPLP-II exhibited a tendency toward moderate values for both the total score and the subscales of health management and nutrition. The physical activity subscale displayed scores close to the upper limit, whereas the spiritual growth and health responsibility dimensions showed distributions tending toward moderate-to-lower scores on their respective scales. On the other hand, anxiety and depression presented a clear tendency toward lower scores, while sleep quality and well-being displayed distributions with moderate-to-high values.

Table 3 presents the bivariate correlations among the five main variables of the study. Well-being showed positive correlations with health-promoting lifestyle (r = 0.377, p < 0.001) and sleep quality (r = 0.408, p < 0.001), whereas it was negatively correlated with anxiety (r = −0.416, p < 0.001) and depression (r = −0.503, p < 0.001). In turn, health-promoting lifestyle was positively associated with sleep quality (r = 0.231, p < 0.001) and negatively associated with anxiety (r = −0.152, p < 0.001) and depression (r = −0.239, p < 0.001). Finally, anxiety and depression were strongly and positively correlated with each other (r = 0.744, p < 0.001), and both were negatively associated with sleep quality (anxiety: r = −0.323, p < 0.001; depression: r = −0.403, p < 0.001).

Table 4 presents the results of the hypothesized and final structural models. The initial tested model showed an unacceptable fit to the data, with TLI values below 0.90 and SRMR and RMSEA values exceeding 0.08. To optimize the overall model fit, modification indices (MIs) were examined, revealing a theoretically justified residual covariance between anxiety (GAD-7) and depression (PHQ-9). Incorporating this covariance substantially improved model fit (MI = 86,186.00; expected parameter change [EPC] = −1.76). After this re-specification, the final model showed an acceptable and substantially improved fit to the data (χ² (1262) = 35,007.519, p < 0.001; CFI = 0.972; TLI = 0.971; SRMR = 0.065; RMSEA = 0.063, 90% CI [0.063, 0.064]).

Figure 2 presents the final structural model, showing the standardized coefficients of the direct effects among the constructs. Within the SEM framework, the second-order latent factor of health-promoting lifestyle showed strong and statistically significant loadings on its first-order factors, ranging from 0.639 to 0.944 (p < 0.001). The results indicated that health-promoting lifestyle negatively predicted anxiety (β = −0.210, [95% CI = −0.236, −0.183], p < 0.001) and depression (β = −0.334, [95% CI = −0.359, −0.309], p < 0.001), and positively predicted sleep quality (β = 0.360, [95% CI = 0.337, 0.383], p < 0.001) and well-being (β = 0.163, [95% CI = 0.134, 0.192], p < 0.001). In turn, depression negatively predicted well-being (β = −0.490, [95% CI = −0.574, −0.432], p < 0.001), whereas sleep quality positively predicted well-being (β = 0.338, [95% CI = 0.316, 0.359], p < 0.001). The association between anxiety and depression was positive and strong (β = 0.856, [95% CI = 0.845, 0.866], p < 0.001). However, the model showed that the direct effects of anxiety on well-being were not statistically significant (β = −0.019, [95% CI = −0.075, 0.036], p = 0.489).

Table 5 presents the standardized specific indirect, path-specific combined, and total effects. Regarding the specific indirect effects, health-promoting lifestyle significantly predicted well-being through depression (β = 0.164, [95% CI = 0.141, 0.186], p < 0.001) and through sleep quality (β = 0.122, [95% CI = 0.111, 0.132], p < 0.001), whereas the pathway through anxiety was not significant (β = 0.004, [95% CI = −0.008, 0.016], p = 0.491). The total indirect effect, considering all mediators jointly, was significant (β = 0.289, [95% CI = 0.269, 0.310], p < 0.001).

The path-specific combined effects on well-being (direct + indirect via each mediator) indicated values of β = 0.326 (95% CI = 0.298, 0.354; p < 0.001) for the depression pathway, β = 0.167 (95% CI = 0.133, 0.200; p < 0.001) for the anxiety pathway, and β = 0.284 (95% CI = 0.259, 0.309; p < 0.001) for the sleep quality pathway. The total effect (direct + all indirect effects) of health-promoting lifestyle on well-being was β = 0.452 (95% CI = 0.429, 0.475; p < 0.001). Regarding the mediated proportion of the total effect, 36.2% of the overall effect was explained through depression (p < 0.001), 26.9% through sleep quality (p < 0.001), and only 0.9% through anxiety, the latter being non-significant (p = 0.490). Finally, the coefficients of determination (R²) indicated that the model accounted for 61.6% of the variance in well-being, 13.0% in sleep quality, 4.4% in anxiety, and 11.2% in depression.